[Noise Figure] Why is the noise figure of some SOAs less than 3 dB

It is widely acknowledged in the industry that the theoretical limit of the noise figure is 3 dB. However, during our actual noise figure testing of SOAs, we frequently encounter noise figure values below 3 dB (typically for SOAs operating under high bias conditions), which often leads customers to question whether our data is falsified or our test methods are improper. This paper analyzes and discusses noise figure testing for high-bias SOAs from multiple perspectives, including the underlying principles of international standards, the characteristics of polarization-maintaining SOA devices, our company’s practical testing specifications, and actual measurement cases, and welcomes feedback and corrections from readers.

## I. Why a 3 dB correction is required when measuring NF using the Polarization Null (PN) method as specified in national standards?
In accordance with IEC 61290-3-1 and the national standard GB/T 16850.3, the core principle of the Polarization Null (PN) method can be summarized in one sentence:
The amplified spontaneous emission (ASE) inside all conventional optical amplifiers exhibits random polarization, with equal noise power in the TE and TM polarization paths. The polarizer in the test optical path filters out the polarization component orthogonal to the signal, and the OSA can only capture half of the ASE power. For this reason, a mandatory 3 dB compensation term is added to the calculation formula to restore the true total ASE noise level of the device.

Excerpts from the original text of IEC 61290-3-1 are shown in the figure below:

Figure 1 Description of noise polarization characteristics in PN method testing specified in international standards
The red-marked section in the figure states that noise is randomly polarized per the standard. Therefore, only half of the ASE noise is captured when testing via the polarization elimination method.

Figure 2 Summary of International Standards: Test Procedures of the PN Method, i.e., Calculation Formula

The standard provides a complete ASE calculation formula for the polarization nulling test procedure:
PASE(total)=Pmeas+Lpol+PCF+3
PASE(total): Corrected total ASE optical power, unit: dBm;
Pmeas: ASE noise measured by an optical spectrum analyzer (OSA), unit: dBm;
Lpol: Insertion loss of the polarization nulling unit used in the test, unit: dB;
PCF: Power correction factor of the OSA, representing the difference in optical power readings between an optical power meter and the OSA;
3: Since the polarization nulling method only captures ASE noise along one polarization axis, a mandatory 3 dB offset is added in the formula to restore the actual total ASE noise level of the device.

However, the +3 dB offset in the formula is subject to explicit physical prerequisites: the tested optical amplifier delivers approximately equal gain for both TE and TM polarizations, and its output contains equal amounts of ASE noise in the two orthogonal polarizations. The 3 dB compensation offsets the half of random polarized ASE filtered out by the polarizer.

Conventional EDFAs / Non-polarization-maintaining SOAs fully satisfy the standard prerequisites:
For polarization-insensitive SOAs, the gain difference between TE and TM modes is less than 1 dB, with ASE power evenly split between the two polarizations;
During PN testing, orthogonally polarized ASE is blocked by the polarizer, leaving the OSA reading at only half of the true noise level;
Without the 3 dB correction, the calculated noise figure (NF) will be consistently 3 dB lower than the actual value, leading to falsely superior performance indicators;
There is a universal industry consensus: a 3 dB polarization correction must be applied for PN testing of non-polarization-maintaining devices.

## II. Should a 3 dB correction be applied in PN-based NF testing of PM-SOAs per the standard?
SOAs are generally categorized into two types by polarization characteristics:
1. Polarization-Insensitive SOA (PI-SOA): Its gain difference between TE and TM modes, also known as Polarization Dependent Gain (PDG), is typically minimal. Devices with PDG below 2 dB are generally classified as PI-SOAs.
2. Polarization-Maintaining SOA (PM-SOA), also referred to as single-polarization or high-polarization SOA: It features a large PDG and a high Polarization Extinction Ratio (PER).

Fundamental difference in ASE polarization noise between PM-SOAs and conventional amplifiers:
Spontaneous emission remains random in origin, yet the asymmetric waveguide of the high-polarization SOA chip introduces extreme loss for TM modes, resulting in nearly no amplification of TM polarized light;
Mass-produced PM-SOAs deliver an output PER of no less than 18 dB, meaning the power of TM polarized ASE accounts for less than 1% of the primary TE polarized ASE and can be approximated as negligible in engineering practice;
The device natively outputs ASE noise of only a single polarization, failing to meet the standard’s prerequisite of equal random noise across two orthogonal polarizations.
Some high-polarization devices integrate a polarization isolator at the internal output port to further boost the PER, rendering TM polarized ASE power far weaker than TE polarized ASE and fully negligible.

In simple terms: Conventional amplifiers inherently generate orthogonal TE and TM noise; half is filtered by the polarizer, necessitating a 3 dB compensation. Polarization-maintaining SOAs only produce TE ASE noise upon manufacture, leaving no orthogonal noise for the polarizer to block, so artificially adding a correction for non-existent noise is theoretically unnecessary. We therefore conclude that the 3 dB correction specified for conventional amplifiers should not be applied.

## III. Internal Operational Specifications of Our Company
Based on the above analysis, our internal rules are defined as follows:
1. Conventional amplifiers (including standard EDFAs and polarization-insensitive SOAs) generate orthogonal TE and TM noise. Half of the noise is filtered by the polarizer, requiring a 3 dB correction.
2. High-polarization optical amplifiers such as PM-SOAs only produce TE ASE noise and shall not be subject to the 3 dB correction. Adding a 3 dB offset would artificially fabricate a non-existent component of TM noise, inflating the measured NF and failing to reflect the intrinsic noise performance of the chip. However, to eliminate data disputes and align test methodology with international standards when providing data to customers, the 3 dB correction shall be applied externally.

Two separate sets of standards are implemented for NF testing of PM-SOAs for internal and external use respectively:
1. Internal R&D and chip mechanism characterization (prioritizing true intrinsic device noise)
For PM-SOAs tested via the PN polarization nulling method, no 3 dB polarization correction is applied, and the NF is calculated directly using raw OSA readings.
Applicability limitations: For internal R&D reference only. Not valid for customer delivery, factory inspection, or third-party certification reports, as it fails to comply with unified international standard calculation protocols.
Purpose: Accurately evaluate on-chip noise and optimize quantum well and waveguide structures without interference from the dual-polarization system correction.

2. Factory test reports, customer delivery documents, third-party inspections, project acceptance (prioritizing compliance)
For PM-SOAs, the full calculation workflow specified in IEC 61290-3-1 is strictly followed. The Lpol offset, PCF offset, and +3 dB polarization correction are incorporated before finalizing NF indicators.
Applicable scenarios: Factory quality inspection, customer delivery documentation, authoritative third-party testing, project acceptance reports.
Purpose: Establish a unified industry evaluation benchmark, match random polarization operating conditions of practical communication systems, eliminate data disputes, and satisfy customer performance specifications.

3. Customer communication
Two sets of data shall be provided to customers: raw single-polarization test values (true intrinsic device performance) and system-equivalent NF values after standard correction (final acceptance indicators for complete systems). The standard source and physical meaning of the 3 dB correction shall be explained to resolve customer concerns that device performance seemingly exceeds the quantum noise limit.

## IV. Practical Test Example
A noise figure test case of a high-power 1550 nm high-polarization SOA manufactured by our company is presented below:
Device Under Test
Model: JSA-BT525G35-PM
Specifications: 1550 nm, 25 dBm maximum output power, 35 dB small-signal gain, polarization-maintaining butterfly-packaged SOA
Test Topology

Figure 3 Noise Figure Test Topology of Polarization Elimination Method in National Standard (Source: IEC 61290-3-1:2003 Figure 1)

However, our company's polarization-maintaining SOA is internally integrated with a polarization isolator, which has the same function as the polarization elimination unit. Therefore, we adopted the simplified topology below for the test.

Figure 4 Simplified Polarization Elimination Method Noise Figure Test Topology of Our Company

Comparative experiments were carried out using the test topology specified by national standards and the simplified test topology developed by our company for PM-SOA respectively, and identical test results were obtained. The adoption of this topology eliminates the need for the test procedure of insertion loss of the Lpol polarization elimination unit.

Test Conditions
An OSA was utilized to measure the NF under the incident 1550 nm optical signal, with operating currents set to 500 mA, 600 mA, 700 mA, 800 mA, 900 mA and 1000 mA, and input optical powers set to -24.87 dBm, -19.88 dBm, -14.85 dBm, -9.84 dBm, -4.82 dBm and -2 dBm.

Test Results
Table 1 NF Data Acquired via OSA (PCF Calculated by the OSA)

It can be seen that most results in Table 1 are less than 3 dB, which represents the actual noise figure of the PM-SOA. (Some data points are <0, which are attributed to test errors and can be ignored)

Table 2 NF data corrected by adding 3 dB in accordance with IEC national standards

Table 2 shows that we manually added a 3 dB correction in accordance with IEC standards to align with current standards, ensure compliance, and prevent disputes over data. It can be observed that with the 3 dB correction applied, the noise figure remains below 6 dB when the input optical power ranges from -20 dBm to 0 dBm.

Figure 5 Relationship between NF and Pin
Figure 5 illustrates the relationship between noise figure (NF), input optical power (Pin), and SOA bias current (I). It can be observed that the noise figure reaches its minimum at an input optical power of approximately -5 dBm, which is consistent with theoretical predictions. When there is no input light to the SOA, a baseline ASE noise exists. As the input optical power gradually increases, the signal light stimulates more carrier transitions, suppressing ASE spontaneous emission light and gradually reducing the NF. With further growth of the signal light, saturation occurs, the output optical power of the signal light remains constant, ASE spontaneous emission light intensifies progressively, and the NF rises accordingly.

Figure 6 An Example of NF Measurement with a Spectrometer
(PM-SOA, polarization cancellation method, input optical power of -2 dBm, SOA operating current of 700 mA)

As shown in Figure 6, this graph presents the output results of the spectrometer when measuring the noise figure (NF) of our SOA model JSA-BT525G35-PM via the aforementioned polarization cancellation method, under the conditions of input optical power of -2 dBm and SOA operating current of 700 mA. The yellow curve in the figure represents the input optical spectrum, while the cyan curve denotes the output optical spectrum. The data collected and analyzed by the spectrometer are displayed at the bottom of the figure. Substituting these data into the formula specified in the national standard (with an additional 3 dB correction term) yields a value of 3.805 dB.

Accordingly, when measuring NF using an Optical Spectrum Analyzer (OSA) and the Polarization Null (PN) method, the actual NF value obtained from the OSA is 0.805 dB. To eliminate disputes over test data and ensure full compliance with standardized testing specifications, an extra 3 dB is added in external test reports in accordance with international standards, resulting in a final value of 3.805 dB.

This article was translated using translation software and may contain inaccuracies. The original Chinese version is available on our company’s official Chinese website:http://www.tj.jhbf.cc/